life-span development of self-esteem and its effects on ... · personality processes and individual...

PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES

Life-Span Development of Self-Esteem and Its Effects onImportant Life Outcomes

Ulrich OrthUniversity of Basel

Richard W. Robins and Keith F. WidamanUniversity of California, Davis

We examined the life-span development of self-esteem and tested whether self-esteem influences thedevelopment of important life outcomes, including relationship satisfaction, job satisfaction, occupationalstatus, salary, positive and negative affect, depression, and physical health. Data came from theLongitudinal Study of Generations. Analyses were based on 5 assessments across a 12-year period of asample of 1,824 individuals ages 16 to 97 years. First, growth curve analyses indicated that self-esteemincreases from adolescence to middle adulthood, reaches a peak at about age 50 years, and then decreasesin old age. Second, cross-lagged regression analyses indicated that self-esteem is best modeled as a causerather than a consequence of life outcomes. Third, growth curve analyses, with self-esteem as atime-varying covariate, suggested that self-esteem has medium-sized effects on life-span trajectories ofaffect and depression, small to medium-sized effects on trajectories of relationship and job satisfaction,a very small effect on the trajectory of health, and no effect on the trajectory of occupational status. Thesefindings replicated across 4 generations of participants—children, parents, grandparents, and theirgreat-grandparents. Together, the results suggest that self-esteem has a significant prospective impact onreal-world life experiences and that high and low self-esteem are not mere epiphenomena of success andfailure in important life domains.

Keywords: self-esteem, life-span development, life outcomes

There is an ongoing debate about whether individuals with highself-esteem have better prospects for their life than individualswith low self-esteem. Whereas some studies suggest that globalself-esteem—a person’s overall evaluation or appraisal of his orher worth—has no important influence on relationship success,economic welfare, and health (e.g., Baumeister, Campbell,Krueger, & Vohs, 2003; Boden, Fergusson, & Horwood, 2008;Krueger, Vohs, & Baumeister, 2008), other studies suggest thatself-esteem has a significant impact on important life outcomes(e.g., Swann, Chang-Schneider, & McClarty, 2007, 2008; Trzesni-ewski et al., 2006). At present, the empirical evidence—which wereview next—is still inconclusive with regard to which, if any, lifedomains are affected by self-esteem. Although previous researchhas identified numerous correlates of self-esteem, including avariety of relationship, work, and health factors, this research doesnot demonstrate that self-esteem actually predicts change in these

correlates. For example, although self-esteem is concurrently cor-related with career success (Judge & Bono, 2001), self-esteemmight not predict increases in career success over time. Whetherself-esteem is a cause or consequence (or both) of important lifeoutcomes is a critical question because a causal effect of self-esteem implies that improving self-esteem would have a beneficialeffect on the outcomes associated with self-esteem. Thus, if self-esteem has a causal effect, then this would suggest that the inter-ventions aimed at increasing self-esteem are worthwhile and likelyto contribute to positive life outcomes and reduce the risk formaladaptive outcomes (for a detailed discussion of this issue, seeBaumeister et al., 2003). In contrast, if self-esteem is simply aconsequence, or epiphenomenon (Seligman, 1993), of positive lifeoutcomes, then efforts to boost self-esteem may produce littleconcrete benefit, either for the individual or for society.

The present research addresses this gap in the literature byexamining effects of self-esteem on life-span trajectories of rela-tionship satisfaction, job satisfaction, occupational status, salary,affect, depression, and health, using data from a large longitudinalstudy of four generations of individuals ages 16 to 97 years.Currently, the field lacks a broad theoretical perspective that couldprovide a framework for the present research. By examining pat-terns of findings across developmental contexts (adolescence toold age), we hope to contribute to building a new, overarchingtheory of the causes and consequences of self-esteem across thelife course.

This article was published Online First September 26, 2011.Ulrich Orth, Department of Psychology, University of Basel, Basel,

Switzerland; Richard W. Robins and Keith F. Widaman, Department ofPsychology, University of California, Davis.

This research was supported by Swiss National Science FoundationGrant PP00P1-123370 to Ulrich Orth.

Correspondence concerning this article should be addressed to UlrichOrth, Department of Psychology, University of Basel, Missionsstrasse 62,4055 Basel, Switzerland. E-mail: [email protected]

Journal of Personality and Social Psychology, 2012, Vol. 102, No. 6, 1271–1288© 2011 American Psychological Association 0022-3514/12/$12.00 DOI: 10.1037/a0025558

1271

In addition, we examined the life-span trajectory of self-esteem,so we can better understand how age-related changes in self-esteem correspond to age changes in life outcomes. The availableevidence suggests that self-esteem follows a quadratic trajectoryfrom adolescence to old age, increasing during young and middleadulthood, reaching a peak at about age 60 and declining in old age(Orth, Trzesniewski, & Robins, 2010). However, this trajectoryhas been documented in only one longitudinal data set (Orth et al.,2010; Shaw, Liang, & Krause, 2010). A meta-analysis of mean-level changes in self-esteem suggests that self-esteem increases inyoung adulthood and does not change after age 30 years; however,few longitudinal studies were available after age 30, possiblylimiting the statistical power of analyses in midlife and old age(Huang, 2010). Thus, further research is needed to replicate thefinding, particularly the old age decline, in other longitudinalsamples.

Relations Between Self-Esteem and Life Outcomes

In this section, we review the available evidence on whetherself-esteem influences relationship satisfaction, job satisfaction,occupational status, salary, affect, depression, and health and,conversely, whether self-esteem is influenced by these variables.Because one goal of the present research is to examine the influ-ence of self-esteem on the life-span trajectories of the outcomes,we also briefly review research on age changes in each of theoutcome variables.

Relationship Satisfaction

Cross-sectional studies suggest that self-esteem is positivelycorrelated with relationship satisfaction (Shackelford, 2001; Voss,Markiewicz, & Doyle, 1999). This positive relation may arisebecause individuals with high self-esteem show more relationship-enhancing behaviors, whereas individuals with low self-esteemshow more dysfunctional, relationship-damaging behaviors. Forexample, individuals with low self-esteem are more sensitive torejection and tend to withdraw and reduce interpersonal closenessfollowing conflicts, thereby undermining satisfaction in close re-lationships (Murray, Holmes, & Griffin, 2000; Murray, Rose,Bellavia, Holmes, & Kusche, 2002). Conversely, relationship sat-isfaction may affect self-esteem. For example, satisfying relation-ships may increase one’s perceived relational value of oneself andmay thereby positively influence self-esteem (cf. Leary &Baumeister, 2000). Consistent with this possibility, Andrews andBrown (1995) found that women who reported becoming closer totheir relationship partner increased in self-esteem over the follow-ing years.

Some studies have examined the life-span trajectory of relation-ship satisfaction. In a meta-analysis, age had a small positive effecton marital satisfaction (Karney & Bradbury, 1995). Consistentwith this finding, Gorchoff, John, and Helson (2008) found thatrelationship satisfaction increased over an 18-year period in mid-dle adulthood. In contrast, Umberson, Williams, Powers, Chen,and Campbell (2005) found that relationship satisfaction declinedover an 8-year interval in all age groups from young adulthood toold age. The results of Gilford and Bengtson (1979) suggest thatpositive indicators of relationship satisfaction show a U-shapedrelation with age, whereas negative indicators of relationship sat-

isfaction decrease over the life course. Other studies mappedrelationship satisfaction on relationship duration (which is animperfect, but useful, proxy for age). Some of these studies re-ported a U-shaped curve for relationship satisfaction. Specifically,relationship satisfaction decreased during the first years of a rela-tionship but increased in later years (Anderson, Russell, &Schumm, 1983; Orbuch, House, Mero, & Webster, 1996), whereasother studies found that relationship satisfaction continuously de-creased across time (e.g., VanLaningham, Johnson, & Amato,2001) or remained stable in long-married couples (Vaillant &Vaillant, 1993). In sum, research published to date provides in-consistent evidence regarding how relationship satisfactionchanges across the life span.

Job Satisfaction

Cross-sectional studies have suggested that self-esteem is pos-itively related to job satisfaction (Judge & Bono, 2001). The fewavailable longitudinal studies have suggested that self-esteem pre-dicts changes in job satisfaction (Judge, Bono, & Locke, 2000;Judge & Hurst, 2008). However, there is a lack of longitudinalstudies that have controlled for the prior level of job satisfactionwhile testing for prospective effects of self-esteem on job satis-faction; moreover, there is a lack of longitudinal studies that havetested for possible effects in the opposite direction, that is, whetherjob satisfaction predicts changes in self-esteem.

With regard to its life-span trajectory, most studies have re-ported that job satisfaction increases continuously across adult-hood (Bernal, Snyder, & McDaniel, 1998; Hunt & Saul, 1975;Janson & Martin, 1982; Kalleberg & Loscocco, 1983; for meta-analyses, see Ng & Feldman, 2010, and Rhodes, 1983). For ex-ample, in a large national probability sample, Bernal, Snyder, andMcDaniel (1998) found a small but significant linear relationbetween age and job satisfaction but found no significant quadraticor cubic trends. However, some studies have found a U-shapedrelation between age and job satisfaction, with the lowest jobsatisfaction in middle adulthood (e.g., Clark, Oswald, & Warr,1996; Hochwarter, Ferris, Perrewe, Witt, & Kiewitz, 2006; Warr,1992). For example, Clark, Oswald, and Warr (1996) examinedage differences in job satisfaction in a large cross-sectional sampleand found a quadratic, U-shaped relation, even when controllingfor multiple covariates, such as gender, education, ethnicity, andhealth. To summarize, the available evidence diverges with regardto the trajectory from young to middle adulthood but consistentlysuggests that job satisfaction increases in the second half of work-ing life.

Occupational Status and Salary

Self-esteem is positively correlated with occupational status(Bachman & O’Malley, 1977; Kammeyer-Mueller, Judge, & Pic-colo, 2008) and salary (Judge, Hurst, & Simon, 2009; Twenge &Campbell, 2002), which are key indicators of socioeconomic sta-tus. Occupational status and salary may influence the individual’sperception of his or her relational value and thereby influenceself-esteem (Leary & Baumeister, 2000). However, it is also plau-sible that high self-esteem helps the individual to attain highereducation, to be more successful at the workplace, and conse-quently to gain higher occupational status and salary. At present,

1272 ORTH, ROBINS, AND WIDAMAN

few longitudinal studies have examined prospective relations be-tween self-esteem and occupational status or salary. Kammeyer-Mueller et al. (2008) found that self-esteem predicted occupationalstatus, even when they controlled for prior level of occupationalstatus. Judge and Hurst (2008) found that positive self-evaluationspredicted higher occupational status and salary from adolescenceto middle adulthood and that the effect became stronger withincreasing age. In the study by Judge et al. (2009), positiveself-evaluations prospectively predicted higher income. Using asample of young adults, Salmela-Aro and Nurmi (2007) found thatself-esteem predicted positive career outcomes 10 years later, forexample, whether participants were employed, had a permanentposition, and had a higher salary (they did not control for previouslevels of outcomes). However, we are not aware of previousstudies that tested prospective effects between self-esteem andoccupational status or salary in both directions (i.e., whether self-esteem predicts occupational status or salary and vice versa) andsimultaneously controlled for prior levels of the variables.

With regard to its life-span trajectory, empirical research sup-ports the view that occupational status and salary increase fromyoung adulthood into midlife and then remain high until retirement(Boyd, 2008; Elstad, 2004; Ganzeboom, De Graaf, Treiman, & deLeeuw, 1992; Hauser, Sheridan, & Warren, 1999; Helson & Soto,2005; Judge & Hurst, 2008; Miech, Eaton, & Liang, 2003; Nam &Boyd, 2004). For example, in a large sample ranging in age from18 to 72 years, occupational status strongly increased in youngadulthood and reached a peak in midlife (Miech et al., 2003).Judge and Hurst (2008) examined a large sample of adolescentsand young adults and found that occupational status and paystrongly increased over the next 25 years.

Affect

Measures of positive and negative affect show strong concurrentcorrelations with self-esteem (e.g., Aspinwall & Taylor, 1992;Joiner, 1995; Watson, Suls, & Haig, 2002). However, to ourknowledge, no previous study has examined prospective relationsbetween affect and self-esteem. Consequently, we do not knowwhether affect predicts self-esteem, self-esteem predicts affect,both affect and self-esteem predict the other, or neither affect norself-esteem predicts the other.

With regard to its life-span trajectory, previous research sug-gests that positive affect remains relatively stable from young tomiddle adulthood (Carstensen, Pasupathi, Mayr, & Nesselroade,2000; Charles, Reynolds, & Gatz, 2001), possibly increasing inadulthood (Helson & Soto, 2005; E. M. Kessler & Staudinger,2009; Löckenhoff, Costa, & Lane, 2008; Mroczek & Kolarz, 1998)and then slightly decreasing in old age (Charles et al., 2001;Kunzmann, Little, & Smith, 2000). In contrast, negative affectdecreases from young to middle adulthood (Gross et al., 1997;Helson & Soto, 2005; E. M. Kessler & Staudinger, 2009; Löck-enhoff et al., 2008; Mroczek & Kolarz, 1998), but the decreaselevels off in old age (Carstensen et al., 2000; Charles et al., 2001;Kunzmann et al., 2000). Reviews and theoretical perspectives onthe life-span development of affect have been provided by Charles(2010), Mroczek (2001), and Scheibe and Carstensen (2010).

Depression

Previous research has suggested that low self-esteem is a risk orvulnerability factor for depression. That is, low self-esteem pro-spectively predicts depression, controlling for prior levels of de-pression (for a review, see Orth, Robins, & Roberts, 2008). Theeffect of low self-esteem on depression holds across short timeintervals (e.g., a few days; Metalsky, Joiner, Hardin, & Abramson,1993; Ralph & Mineka, 1998; Roberts & Monroe, 1992) andacross long intervals (e.g., years; Orth et al., 2008; Trzesniewski etal., 2006). Moreover, the effect of low self-esteem on depressionholds for men and women and across all age groups from adoles-cence to old age (Orth, Robins, Trzesniewski, Maes, & Schmitt,2009). In addition, most prior research has failed to support theopposite direction of the relation—that low self-esteem is a con-sequence, rather than a cause, of depression (Ormel, Oldehinkel, &Vollebergh, 2004; Orth et al., 2008; Orth, Robins, Trzesniewski, etal., 2009; but see Shahar & Davidson, 2003; Shahar & Henrich,2010).

With regard to its life-span trajectory, most studies have foundthat depression decreases from young adulthood to middle adult-hood and then increases again in old age, with converging evi-dence provided by both cross-sectional studies (Gatz & Hurwicz,1990; R. C. Kessler, Foster, Webster, & House, 1992; Lewinsohn,Rohde, Seeley, & Fischer, 1991; Miech & Shanahan, 2000; Mi-rowsky & Ross, 1992; but see Blanchflower & Oswald, 2008) andlongitudinal studies (Davey, Halverson, Zonderman, & Costa,2004; Kasen, Cohen, Chen, & Castille, 2003; Mirowsky & Kim,2007; Mirowsky & Reynolds, 2000; Rothermund & Brandtstadter,2003; Wallace & O’Hara, 1992). For example, R. C. Kessler et al.(1992) examined two large nationally representative surveys,which included participants from young adulthood to old age, andfound that age showed a U-shaped quadratic relation with depres-sion, with the lowest depression levels at about age 50 years inboth samples. In a longitudinal study of older adults ranging fromages 60 to 96 years depression consistently increased with age(Davey et al., 2004).

Health

Individuals with high self-esteem tend to report better physicalhealth (e.g., Benyamini, Leventhal, & Leventhal, 2004; Makikan-gas, Kinnunen, & Feldt, 2004). Individuals with high self-esteemmay seek and receive more social support, experience less stress,and show more adaptive coping behaviors, thereby enhancing theirhealth. However, it is also plausible that healthy individuals feelmore independent and better able to contribute to their family andsociety, which, in turn, would bolster self-esteem. Available lon-gitudinal data suggest that self-esteem prospectively predictshealth outcomes. For example, Trzesniewski et al. (2006) foundthat low self-esteem in adolescence predicted more physical healthproblems at age 26 (see also Christie-Mizell, Ida, & Keith, 2010;Stinson et al., 2008). However, only one study has examinedwhether health predicts changes in self-esteem; Reitzes andMutran (2006) found that self-esteem and functional health werereciprocally related across a 2-year interval, controlling for priorlevels in the constructs.

With regard to its life-span trajectory, most studies have sug-gested that physical health stays relatively constant, or only de-

1273SELF-ESTEEM DEVELOPMENT AND LIFE OUTCOMES

clines slowly, from adolescence to the 40s or 50s and then worsensat an accelerating rate into old age. In longitudinal studies, many ofwhich used national probability samples, researchers have examinedthe trajectories of self-rated health (e.g., Liang et al., 2005; Mc-Cullough & Laurenceau, 2005; Ross & Wu, 1996; Sacker, Worts, &McDonough, 2011), impairments of functional health, such as beingconfined to bed or having difficulties with bathing, dressing, climbingstairs, or doing heavy housework (e.g., Chiu & Wray, 2010; J. Kim &Miech, 2009; Liang et al., 2003; Ross & Wu, 1996), physical symp-toms (e.g., Aldwin, Spiro, Levenson, & Cupertino, 2001), and numberof health problems, such as arthritis, hypertension, stroke, or cancer(e.g., House et al., 1994).

The Present Research

The goal of the present research was to examine the life-spandevelopment of self-esteem and its effects on important life out-comes. We used data from a large longitudinal study of individualsranging in age from adolescence to old age. We examined theeffects of self-esteem on the development of relationship satisfac-tion, job satisfaction, occupational status, salary, positive andnegative affect, depression, and health. The cohort-sequential de-sign, spanning four generations of the same families, allowed us todetermine whether any observed changes were due to intraindi-vidual change or cohort differences (Baltes, Cornelius, & Nessel-roade, 1979).

In the first part of this study, we attempted to replicate previousresearch suggesting that self-esteem increases across adulthoodand then decreases in old age (Orth et al., 2010). We tested whichof several growth curve models (intercept only, linear, quadratic,and cubic) yields the best fit to the data. We also used the uniquedesign of the study to test whether the self-esteem trajectory variedacross generations—children, parents, grandparents, and great-grandparents. Finally, we examined whether the trajectory variedas a function of demographic variables (i.e., gender and educa-tion). In the second part of this study, we examined reciprocalprospective relations between self-esteem and life outcomes, usingcross-lagged regression analyses. We tested whether any observedrelations held across generations and across the life span. Finally,in the third part of this study, we examined the influence ofself-esteem on the life-span trajectories of the outcome variables,using growth curve models with self-esteem as a time-varyingcovariate (TVC). For each outcome variable, we also testedwhether the trajectory varied across generations.

As described earlier, we used two types of models (i.e., cross-lagged regression models and growth curve models with a TVC) toexamine the effects of self-esteem on life outcomes. Each type ofmodel yields important information that is not provided by theother type of model. The cross-lagged model tests the direction ofeffects between constructs (which is not provided by the growthcurve model with a TVC), because effects are prospectively testedand autoregressive effects are controlled for. Cross-lagged regres-sion models are, at present, the most frequently used and recom-mended models to test whether data are consistent with causalhypotheses on the relation between constructs, when only nonex-perimental longitudinal data are available (Cole & Maxwell, 2003;Finkel, 1995; Little, Preacher, Selig, & Card, 2007). In contrast,growth curve analyses provide a way to model the developmentaltrajectory of a variable and to examine effects of other variables on

the trajectory (which is not provided by cross-lagged models). Forexample, it is possible that the cross-lagged regression analysisshows that self-esteem predicts relative increases in an outcomevariable, whereas the growth curve analysis shows that a strongnegative developmental trend in the outcome outweighs any ben-eficial effects of self-esteem. Moreover, growth curve models witha TVC also demonstrate whether the average trajectory of theoutcome is altered when the TVC is controlled for, which cannotbe examined with cross-lagged regression models (Bollen & Cur-ran, 2006; Grimm, 2007; Preacher, Wichman, MacCallum, &Briggs, 2008).

The present study extends previous research in several ways.First, the study examines the development of self-esteem across asignificantly broader age range (i.e., adolescence to old age) thanin previous longitudinal studies, providing a more comprehensivepicture of life-span development of self-esteem. Second, the studysystematically tests for cross-lagged prospective effects betweenself-esteem and life outcomes, which helps to clarify the role ofself-esteem as a potential cause or consequence of importantoutcomes. For most of the outcomes examined in this research(i.e., relationship satisfaction, job satisfaction, occupational status,salary, and positive and negative affect), no previous study hastested the reciprocal prospective effects in both directions, con-trolling for prior levels of both constructs. Third, the study exam-ines the influence of self-esteem on the life-span trajectories ofthese outcome variables, allowing us to examine whether thetrajectory of the life outcome is altered when self-esteem is heldconstant and to compare the size of the self-esteem effects tonormative changes in the development of life outcomes. Fourth,the unique design of the study allowed us to test whether theobserved effects replicated across four generations of the samefamilies, providing important information about the generalizabil-ity of the findings across generational cohorts. If the results showthat the hypothesized effects of self-esteem on life outcomes arenot confined to specific cohorts but hold across generations, thefindings have important implications for theory and research onthe life-span consequences of self-esteem.

Method

The data come from the Longitudinal Study of Generations(LSG; Bengtson, 2009). In 1971, three-generation families wererandomly drawn from a subscriber list of about 840,000 membersof a health maintenance organization in Southern California. Since1991, the study has included a fourth generation (i.e., the great-grandchildren in the same families). The members of the healthmaintenance organization included primarily White working-classand middle-class families, and very low and very high socioeco-nomic levels were not represented in the population. However,level of education among family members corresponded to na-tional norms at the time the sample was drawn (Bengtson, Biblarz,& Roberts, 2002). Although the sample was originally recruited inSouthern California, at recent waves, more than half of the samplelived outside the region in other parts of California, in other statesof the United States or abroad, because of residential mobility ofparticipants (Bengtson et al., 2002).

Participants were assessed in 1971, 1985, 1988, 1991, 1994,1997, and 2000. In 1971 and 1985, the LSG did not include the fullself-esteem measure; the present study therefore examines data of


the five waves from 1988 to 2000. We excluded any participantwhose age was unknown or who did not provide data on self-esteem at any of the five waves.

Participants

The sample included 1,824 individuals (57% female). Table 1gives an overview of the demographic characteristics for the fullsample and for the four separate generations. The distribution ofgender is relatively even across generations. The age range acrosswaves was 14 to 102 years; however, because only one assessmentwas below age 16 and two assessments were above age 97, werestricted the analyses to the age range from 16 to 97 years. Of theparticipants, 94% were Caucasian, 3% were Hispanic, 1% wereAfrican American, 1% were Native American, and 1% were ofother ethnicity. Because of the low frequencies of ethnicities otherthan Caucasian, we did not examine ethnic differences.

Data on study variables were available for 1,448 individuals in1988, for 1,463 individuals in 1991, for 1,405 individuals in 1994,for 1,281 individuals in 1997, and for 1,227 individuals in 2000.To investigate the potential impact of attrition, we comparedindividuals who did and did not participate in the most recent waveof data collection (2000) on study variables assessed at the firstwave (1988). Participants who dropped out versus those who didnot were more likely to be older (Ms � 52.1 vs. 43.6 years,respectively; d � 0.48), were less likely to be female (53% vs.60%, respectively), had lower levels of education (Ms � 4.55 vs.5.26, respectively; d � �0.45), and reported slightly lower self-esteem (Ms � 3.45 vs. 3.51, respectively; d � �0.13), lesspositive affect (Ms � 0.74 vs. 0.79, respectively; d � �0.19), lessnegative affect (Ms � 0.31 vs. 0.35, respectively; d � �0.13),more depression (Ms � 1.53 vs. 1.47, respectively; d � 0.13), andpoorer physical health (Ms � 2.97 vs. 3.18, respectively; d ��0.28); differences in relationship satisfaction, job satisfaction,occupational status, and salary were all nonsignificant. Althoughdifferences in demographic variables were of medium size, differ-ences in self-esteem and the life outcome variables were small tononsignificant. Thus, nonrepresentativeness because of attritionwas not a serious concern in the present study.

Measures

Self-esteem. Self-esteem was assessed with the 10-itemRosenberg Self-Esteem Scale (RSE; Rosenberg, 1965), the most

commonly used and well-validated measure of self-esteem (Rob-ins, Hendin, & Trzesniewski, 2001). At all waves, responses weremeasured with a 4-point scale, ranging from 1 (strongly disagree)to 4 (strongly agree). However, in 1988 and 1991, the scale labelsof the middle response categories (i.e., 2 � somewhat disagree and3 � somewhat agree) differed slightly from the labels used in1994, 1997, and 2000 (i.e., 2 � disagree and 3 � agree). Conse-quently, the raw RSE scores (e.g., the means of the RSE items)were not equivalent across assessments with differing labels andcould not be used for growth curve analyses. Therefore, we com-puted RSE factor scores by using confirmatory factor analysis withcategorical indicators and equating by common items (these pre-paratory analyses are reported at the end of the Method section).The alpha reliability of the RSE was .86 in 1988, .83 in 1991, .86in 1994, .86 in 1997, and .86 in 2000.

Relationship satisfaction. The LSG uses a 10-item relation-ship satisfaction scale (Gilford & Bengtson, 1979). Example itemsare “you calmly discuss something together,” “you laugh to-gether,” “you disagree about something important” (reversescored), and “one of you becomes critical and belittling” (reversescored). Participants reported how frequently they experiencedthe situation when they were with their spouse or partner on a5-point scale (1 � hardly ever; 2 � sometimes; 3 � fairly often;4 � quite frequently; 5 � almost always), with M � 3.90 (SD �0.70) averaged across the five waves. The alpha reliability was .87in 1988, .89 in 1991, .87 in 1994, .89 in 1997, and .88 in 2000.

Job satisfaction. Job satisfaction was assessed using a singleitem: “How satisfied would you say you are with your main job?”Responses were measured on a 5-point scale (1 � not at allsatisfied; 2 � not too satisfied; 3 � somewhat satisfied; 4 � verysatisfied; 5 � extremely satisfied), with M � 3.63 (SD � 0.92)averaged across the five waves.

Occupational status. The LSG provides a measure of occu-pational status created from the Hauser–Warren SocioeconomicIndex (Hauser & Warren, 1997), which is based on the averageeducation level and income of persons in each occupation asreflected in the 1990 U.S. Census. Participants’ scores are based oninformation for their current or most recent occupation. Examplesfor occupational status scores are 14.9 for farm workers, 29.8 formining machine operators, 50.2 for librarians, 64.1 for teachers insecondary schools, and 77.1 for dentists (for further information,see Bengtson, 2009, Appendix C1). The measure ranged from 9.6to 80.5 and had a mean of 43.5 (SD � 14.6) averaged across thefive waves.

Salary. Salary was assessed with a 12-point measure, rangingfrom 1 (less than $10,000) to 12 ($110,000 or more). The meanwas 4.07 (SD � 2.55) averaged across the five waves.

Positive and negative affect. Positive and negative affectwere assessed with the Affect Balance Scale (Bradburn, 1969),with five items measuring positive affect and five items measuringnegative affect. The validity of the scale has been repeatedlysupported (Harding, 1982; K. A. Kim & Mueller, 2001). Exampleitems are “particularly excited or interested in something” and“that things were really going your way” (positive affect) and “sorestless that you couldn’t sit long in a chair” and “bored” (negativeaffect). For each item, participants reported whether they ever feltthat way during the past few weeks (0 � no; 1 � yes). For positiveaffect, the mean was 0.76 (SD � 0.26), and for negative affect, themean was 0.32 (SD � 0.30), averaged across the five waves. In

Table 1Demographic Characteristics of the Sample

Sample NFemale

(proportion)Mean age (and SD)

in 1991aAge range

across waves

Generation 1 176 .64 83.0 (5.3) 61–97Generation 2 601 .57 63.3 (5.2) 41–90Generation 3 851 .56 39.2 (2.7) 19–68Generation 4 196 .59 20.1 (4.0) 16–47Full sample 1,824 .57 49.3 (18.3) 16–97

Note. Age is reported in years. The age range is given for observationswith data on study variables.a Mean age is reported for 1991 (i.e., the second wave) because Generation4 individuals were not part of the study in 1988 (i.e., the first wave).


particular, when items are dichotomous, coefficient alpha canunderestimate the reliability of scales. We therefore used themethod by Raykov, Dimitrov, and Asparouhov (2010) for estimat-ing the reliability of scales with dichotomous items. The estimatesfor positive affect were .84 in 1988, .87 in 1991, .84 in 1994, .87in 1997, and .85 in 2000, and the estimates for negative affect were.69 in 1988, .69 in 1991, .72 in 1994, .71 in 1997, and .70 in 2000.

Depression. Depression was assessed with the 20-item Cen-ter for Epidemiologic Studies Depression Scale (CES-D; Radloff,1977). The CES-D is a frequently used and well-validated measurefor the assessment of depressive symptoms in nonclinical, subclin-ical, and clinical populations (Eaton, Smith, Ybarra, Muntaner, &Tien, 2004). For each item, participants reported how frequentlythey experienced the symptom during the past week, using a4-point scale (0 � rarely or none of the time; 1 � a little of thetime; 2 � a moderate amount of the time; 3 � most or all of thetime). Because two of the CES-D items were used for equating theRSE across assessments (further details provided later), theseitems were excluded from the CES-D; thus, we used an 18-itemCES-D for the analyses on depression. At each wave, the 18-itemCES-D correlated .99 with the full 20-item CES-D. The 18-itemCES-D had a mean of 1.52 (SD � 0.44) averaged across the fivewaves and had an alpha reliability of .88 in 1988, .91 in 1991, .90in 1994, .90 in 1997, and .90 in 2000.

Health. Participants rated their health on a single item:“Compared to people of your own age, how would you rate youroverall physical health at the present time?” Responses weremeasured on a 4-point scale (1 � poor; 2 � fair; 3 � good; 4 �excellent), with M � 3.07 (SD � 0.76) averaged across the fivewaves.

Health problems. Health problems were assessed with anindex of 10 dichotomous items. The items were heart problems/angina, high blood pressure, stroke, cancer, respiratory ailments,digestive problems, arthritis/rheumatism, diabetes, cataracts/glaucoma/retinal degeneration, and hearing impairment. The meanwas 1.29 (SD � 1.55) averaged across the five waves. Because thedistribution of health problems is positively skewed, with thelargest frequency at zero problems, the variable was used inlogarithmic metric.1

Education. The LSG includes an 8-point measure of educa-tion (1 � 8th grade or less; 2 � some high school, 9th–11th grade;3 � high school or vocational school graduate; 4 � specializedtechnical, business, or other training after high school; 5 � somecollege, 1–3 years; 6 � college or university graduate; 7 � one ormore academic years beyond college, including MA; 8 � post-graduate degree, PhD, MD, JD, etc.). For most participants, be-cause of their age, level of education was invariant across waves,so we used the highest degree that participants reported acrosswaves as a time-invariant covariate in our analyses. The mean was5.04 (SD � 1.62).

Statistical Analyses

Analyses were conducted with the Mplus 6.1 program (Muthen& Muthen, 2010). To deal with missing values, we used fullinformation maximum likelihood estimation to fit models directlyto the raw data, which produced less biased and more reliableresults compared with conventional methods of dealing with miss-

ing data, such as listwise or pairwise deletion (Allison, 2003;Schafer & Graham, 2002; Widaman, 2006).

For confirmatory factor models and cross-lagged regressionmodels, fit was assessed by the comparative fit index (CFI), theTucker–Lewis index (TLI), and the root-mean-square error ofapproximation (RMSEA), based on the recommendations of Huand Bentler (1999) and MacCallum and Austin (2000). Hu andBentler suggested that good fit is indicated by values greater thanor equal to .95 for CFI and TLI and by values less than or equal to.06 for RMSEA. To test for differences in model fit, we used thetest of small differences in fit recommended by MacCallum,Browne, and Cai (2006, Program C). For these tests, statisticalpower was high with values above .99 (MacCallum et al., 2006,Program D).

For growth curve models, CFI, TLI, and RMSEA were notavailable; therefore, the fit of these models was assessed with theBayesian information criterion (BIC). For BIC, absolute valuescannot be interpreted, but when comparing models, lower valuesindicate better model fit.

Equating the RSE Across Waves

In the LSG, the labels of the RSE response categories were alteredin 1994 (as described earlier), which likely precludes measurementinvariance of the RSE raw scores across waves. Measurement invari-ance is, however, a basic requirement for growth modeling becausescores at different waves are directly comparable only when measure-ment invariance holds (Edwards & Wirth, 2009; Widaman, Ferrer, &Conger, 2010). To resolve this problem, we used confirmatory factoranalysis with categorical indicators (Wirth & Edwards, 2007) andequating by common items (Edwards & Wirth, 2009) to computeRSE factor scores. This procedure can establish measurement invari-ance because all observed scores, by using the common items asanchors, are mapped on the same latent scale.

To equate by common items, we needed items that were concep-tually related to self-esteem and that were available at each wave inidentical response format. Two CES-D items met these requirements.The items were “I felt that I was just as good as other people” (Item4 of the CES-D) and “I thought my life had been a failure” (reversescored; Item 9 of the CES-D). The content of these items suggests thatthey are essentially measures of self-esteem. Thus, the self-esteemmeasurement models included 10 RSE items and two CES-D items;all items were available at each wave. For the RSE items, twodifferent sets of labels for response categories were used (i.e., Set Ain 1988 and 1991 and Set B in 1994, 1997, and 2000). For the CES-Ditems, an identical set of labels was used across waves (i.e., Set C). Ina series of increasingly restricted measurement models, we tested themeasurement invariance of the CES-D items. If measurement invari-ance of the CES-D items holds, then the RSE items can be mapped onthe same scale across waves. In addition, we also tested the measure-ment invariance of the RSE items for waves using the same set oflabels (i.e., Set A and Set B). Although measurement invariance of theRSE items is not required for equating by the CES-D items, its

1 We did not compute coefficient alpha for the index of health problems.Coefficient alpha is not an appropriate measure of reliability for this scalebecause health problems are an emergent not latent construct, defined by anaggregation of relatively independent indicators (Bollen & Lennox, 1991;Streiner, 2003).


presence would strengthen confidence in the measurement propertiesof the RSE.2

The measurement models included five correlated self-esteem fac-tors (one factor per wave). In addition, the models included methodfactors that accounted for bias due to positive and negative wording ofthe items (cf. Marsh, Scalas, & Nagengast, 2010). Both the positiveand negative wording factors were correlated across waves, but pos-itive wording factors were uncorrelated with negative wording fac-tors, and all wording factors were uncorrelated with the self-esteemfactors. Including these method factors strongly increased the fit of themodels. Also, the measurement models included longitudinal corre-lations between the same items measured at different waves (Cole &Maxwell, 2003). Including these correlations controls for possiblebias due to indicator-specific variance that is not captured by theself-esteem and wording factors. We analyzed the indicators as cate-gorical variables, using the mean- and variance-adjusted weightedleast squares (WLSMV) estimator.

The first measurement model (i.e., Model 1) included configuralinvariance (Widaman et al., 2010) for both the CES-D and RSEitems. The fit of this model was good (Table 2). Models 2 and 3tested weak and strong invariance of the CES-D items by progres-sively constraining the loadings and thresholds, respectively, to beequal across waves. The fit of these models was as good as the fitof Model 1, supporting the conclusion that the CES-D items showmeasurement invariance and can be used for equating the RSEfactor scores across waves.3 Models 4 and 5 tested the measure-ment invariance of the RSE items for waves using the same set oflabels (i.e., Set A and Set B), again by progressively constrainingthe loadings and thresholds, respectively, to be equal. The fit ofthese models was good (Table 2). Although the RMSEA values

slightly worsened, the difference was small, and the CFI and TLIvalues were unaltered. We concluded that, when using the sameresponse format, the RSE items showed measurement invariance.4

The RSE factor scores were computed in Model 6, which includedonly the RSE items and omitted the CES-D items. The loadingsand thresholds of the RSE items were fixed to the parameterestimates from Model 5. The participants’ scores on the five latentself-esteem factors were saved and used as measures of self-esteem in the remainder of the analyses.

Results

Developmental Trajectory of Self-Esteem

We examined the trajectory of self-esteem from adolescence to oldage, by using growth curve models that capture the development ofself-esteem across the entire observed age range represented in thesample. Although each participant provided data for, at most, five agepoints (covering a 12-year interval), the complete life-span trajectorywas constructed using information from all participants simultane-ously. This approach is based on the assumption, which is tested later,that a common trajectory can be modeled across all generationsincluded in the sample (e.g., Duncan, Duncan, & Strycker, 2006;Preacher et al., 2008). To account for the fact that the measurementwas asynchronous across age (i.e., the data are organized by waves,but we were interested in another metric of time, specifically theindividuals’ age at each wave), we used individual slope loadings,following the recommendations by Mehta and West (2000), Bollenand Curran (2006), and Preacher et al. (2008).

We estimated a model with an intercept only and linear, qua-dratic, and cubic growth models (Preacher et al., 2008). Becausethe slope loadings are based on age rather than the five measure-ment occasions, it was possible to estimate relatively complextrajectories. Age was centered at 50 years. The cubic model hadthe best fit to the data (the BIC values were 13,483.5 for the modelwith an intercept only; 13,429.4 for the linear model; 13,037.0 forthe quadratic model; and 12,906.6 for the cubic model). Therefore,

2 Although the LSG assessments in 1971 and 1985 (i.e., the two wavesthat are not used in the present research) included some items of the RSE,we did not use these data for the following reasons. In 1971, the LSG usedlabels for the RSE response categories that differed from both sets of labelsused in the assessments from 1988 to 2000. Moreover, in 1971 the CES-Dwas not included; therefore, no common items were available to equate theRSE scores from 1971 with scores from any other waves. In 1985, only oneRSE item was used, so the reliability and validity of RSE factor scores for1985 would have depended on a single item and likely would have beenlow. We therefore decided against using data from the 1971 and 1985assessments of the LSG.

3 Because the chi-square value for the WLSMV cannot be used formodel comparison, we did not formally test for differences in fit of themeasurement models but examined the fit indicators CFI, TLI, andRMSEA.

4 We also tested the fit of a model that imposed measurement invarianceof the RSE items across all five waves, disregarding the alteration in thelabels of response categories. As expected, the fit of this model (withCFI � .90, TLI � .98, and RMSEA � .051) was clearly worse than the fitof the measurement invariance models that accounted for the alteration inthe labels. This result confirms that the difference in response formatsneeds to be taken into account in the measurement models.

Table 2Fit of Self-Esteem Measurement Models With CategoricalIndicators and Equating by Common Items

Model CFI TLI RMSEA

Invariance of CES-D items (with configuralinvariance of RSE)

1. Configural invariance .94 .99 .0342. Weak invariance .95 .99 .0333. Strong invariance .95 .99 .033

Invariance of RSE (with strong invariance ofCES-D items)

4. Weak invariance .95 .99 .0355. Strong invariance .95 .99 .036

Computation of RSE factor scores (with CES-Ditems excluded)

6. Strong invariancea .98 .99 .032

Note. In the LSG, the labels of the RSE response categories were alteredin 1994. Therefore, in a series of measurement models, RSE factor scoreswere equated by using two CES-D items that are essentially measures ofself-esteem and that were available at each wave in identical responseformat (see text for further explanation). Weak invariance � equalityconstraints on loadings; strong invariance � equality constraints on load-ings and thresholds. For models with categorical indicators, confidenceintervals of RMSEA are not available. CFI � comparative fit index; TLI �Tucker–Lewis index; RMSEA � root-mean-square error of approxima-tion; CES-D � Center for Epidemiologic Studies Depression Scale;RSE � Rosenberg Self-Esteem Scale; LSG � Longitudinal Study ofGenerations.a Parameters fixed to estimates of Model 5.


in the remainder of the analyses, we estimated a cubic self-esteemtrajectory. Next, we tested whether there are differences betweengenerations in the trajectory. Using a multiple-group analysis, wetested whether a model in which coefficients are freely estimatedyielded a better fit than a model with cross-generation equalityconstraints on the coefficients.5 In these models, the variance ofthe cubic growth factor had to be set to zero to allow for conver-gence of the models. The results showed that a model with con-straints forcing the same trajectory across generations fit betterthan a model without the constraints, suggesting there are nogenerational differences in the self-esteem trajectory.

Thus, the evidence suggests that modeling a single coherent tra-jectory across the observed age range is appropriate. Figure 1A showsthe average predicted trajectory of self-esteem for the full sample.Overall, self-esteem tended to increase during adolescence, youngadulthood, and middle adulthood, reached a peak at age 51, and thendeclined in old age. There was about a one quarter standard deviationincrease (d � 0.29) from age 16 to 51 and about a two third standarddeviation decrease (d � �0.67) from age 51 to 97 years. Although thetrajectory shown in Figure 1A appears to be quadratic, it is in factcubic. The cubic factor results in a lower increase in young adulthoodand a stronger decrease in old age than would be explained by aquadratic trajectory. Although one of the turning points of the cubictrajectory is outside of the observed age range, the cubic factor isimportant to fit the trajectory more closely to the data (as indicated bythe better model fit of the cubic trajectory).

Next, we estimated conditional growth curve models (Preacheret al., 2008) to examine the moderating effects of gender andeducation on the self-esteem trajectory. Gender was dummy codedfor these analyses, and education was converted to z scores. Forboth gender and education, none of the effects on the intercept oron the linear, quadratic, and cubic growth factors were significant,except that education significantly predicted the intercept (theunstandardized regression coefficient was 0.17; no standardizedestimate is available). We also examined the effects of gender andeducation simultaneously: The effects of gender remained nonsig-nificant, and the education effect on the intercept remained signif-icant at 0.16. Figure 1B illustrates the education effect by plottingthe predicted self-esteem trajectory for individuals with high (i.e.,one standard-deviation unit above the mean) and low (i.e., onestandard-deviation unit below the mean) levels of education. Ascan be seen, participants with higher levels of education had higherself-esteem at all ages: The self-esteem difference between partic-ipants with low versus high education at age 16 corresponded tod � 0.38 and at age 97 corresponded to d � 0.24.

Cross-Lagged Effects of Self-Esteem andLife Outcomes

Before we examined the effects of self-esteem on the life-spantrajectories of the outcome variables, we first tested whether self-esteem in fact predicts the outcome variables by using cross-lagged regression models. Figure 2 provides a generic illustrationof these models (cf. Finkel, 1995). Each model tested the relationsbetween self-esteem and one of the outcome variables.6 The rela-tions between variables were specified as cross-lagged effects,which indicates the prospective effect of one variable on the other(e.g., the effect of self-esteem in 1988 on relationship satisfactionin 1991), after controlling for their stability across time (e.g., theeffect of relationship satisfaction in 1988 on relationship satisfac-tion in 1991). We accounted for variance due to specific measure-ment occasions by correlating the residual variances within waves(e.g., the residual of self-esteem in 1991 and the residual ofrelationship satisfaction in 1991; cf. Cole & Maxwell, 2003). Thestability and cross-lagged coefficients were constrained to be equalacross time. Although the fit values were slightly worse than thenormative values specified by Hu and Bentler (1999), we judgedthe fit of the models to be overall satisfactory: The CFI valuesranged from .93 to .97, the TLI values ranged from .92 to .96, andthe RMSEA values ranged from .065 to .095. For all outcome

5 In tests of differences between generations, participants from Genera-tion 4 could not be included because for one of the waves no data wereavailable (as reported in the Method section, this generation did notparticipate in the 1988 assessment). Moreover, in tests of generationaldifferences for the variables job satisfaction, occupational status, andsalary, participants from Generation 1 could not be included because thevariables were not assessed for this generation (by reasons of study design).

6 The cross-lagged regression analyses for self-esteem and depressionare mostly redundant with analyses reported in a previous study, whichused data from the 1988 to 1997 assessments of the LSG (Orth, Robins,Trzesniewski, et al., 2009, Study 1). In the present article, we report thecross-lagged regression analyses for depression for reasons of complete-ness; however, the growth curve analyses of depression reported in theremainder of the article have not been published previously.

Figure 1. Average predicted trajectory of self-esteem for the full sample(Panel A) and for individuals with high (i.e., one standard-deviation unitabove the mean) and low (i.e., one standard-deviation unit below the mean)levels of education (Panel B). Measures were converted to z scores for theanalysis.


variables, the equality constraints on the stability and cross-laggedcoefficients did not significantly worsen model fit. Using multiple-group models, we also tested for generational differences in thestability and cross-lagged coefficients. Importantly, models withcross-generation equality constraints did not fit significantly worsethan models in which the coefficients were estimated freely.

Table 3 reports the estimates for the stability and cross-laggedcoefficients.7 The cross-lagged effects showed a consistent picture:Self-esteem prospectively predicted each of the outcome variables,whereas the outcome variables generally did not prospectivelypredict self-esteem. In all cases, the effect of self-esteem on theoutcome variable was larger than the effect of the outcome vari-able on self-esteem. The results indicate that individuals with highself-esteem subsequently reported higher levels of relationshipsatisfaction, job satisfaction, occupational status, salary, positiveaffect, and health and reported lower levels of negative affect,depression, and health problems. The multigroup analyses de-scribed earlier suggest that these effects held across all four gen-erations. In sum, the results suggest that it is appropriate toexamine the influence of self-esteem on life-span trajectories ofthe outcome variables rather than vice versa.

Effect of Self-Esteem on Developmental Trajectories ofLife Outcomes

To establish a baseline trajectory, we first modeled the basictrajectories of each outcome variable (i.e., without taking self-esteem into account).8 As in the growth models for self-esteem, weused individual slope loadings, based on age, to model the trajec-tory across the observed age range using information from allparticipants simultaneously. For each outcome, we estimated amodel with an intercept only and linear, quadratic, and cubicgrowth models. For the analyses, age was centered at 50 years, andmeasures were converted to z scores.9 For salary and health prob-lems, multigroup analyses indicated statistically significant gener-ational differences in their trajectories; therefore, it was not ad-missible to model a single life-span trajectory for these outcomes.Consequently, we did not include these outcomes in the growthcurve analyses with self-esteem as a TVC. For all other lifeoutcomes, however, there were no significant generational differ-ences in their trajectory. For job satisfaction and positive affect,the linear model had the best fit; for relationship satisfaction andnegative affect, the quadratic model had the best fit; and foroccupational status, depression, and health, the cubic model hadthe best fit.

We next examined the effect of self-esteem on the trajectories,by using growth curve models with a TVC (see Bollen & Curran,2006; Preacher et al., 2008). Figure 3 shows a generic illustrationof the models, for the case of quadratic growth in the outcomevariable (the models for linear and cubic growth were specifiedaccordingly). Again, the models included individual slope loadings

7 The standardized coefficients were averaged across time intervalsusing Fisher’s Zr transformations. Although the coefficients were con-strained to be equal across time intervals, the constraints were imposed onunstandardized coefficients (as typically recommended), which led toslight variation in the resulting standardized coefficients.

8 A previous study examined growth in positive and negative affectusing data from the LSG (Charles et al., 2001). However, the study did notinclude any analyses of self-esteem or, specifically, of the influence ofself-esteem on the trajectories of positive and negative affect.

9 For job satisfaction and occupational status, figures show predictedtrajectories from age 16 to age 70 because only a few observations wereavailable outside of this range.

Figure 2. Cross-lagged regression model of the relations between self-esteem and an outcome variable (Y). Therelations between factors are specified as cross-lagged effects, which indicate the prospective effect of onevariable on the other (e.g., the effect of self-esteem in 1988 on Y in 1991), after controlling for their stabilityacross time (e.g., the effect of Y in 1988 on Y in 1991). Residual variances (i.e., disturbances) are denoted as d1,d2, and so forth.

Table 3Cross-Lagged and Stability Effects of Self-Esteem andOutcome Variables

Outcome variable (Y)

Cross-lagged effects Stability effects

SE 3 Y Y 3 SE SE 3 SE Y 3 Y

Relationship satisfaction .05� .01 .85� .72�

Job satisfaction .14� �.01 .85� .39�

Occupational status .03� .00 .85� .76�

Salary .03� .01 .85� .83�

Positive affect .17� .00 .85� .43�

Negative affect �.13� .02� .86� .51�

Depression �.20� .02� .86� .46�

Health .11� .02� .85� .59�

Health problems �.05� �.02� .85� .71�

Note. The table shows standardized regression coefficients. SE � self-esteem.� p � .05.


based on age. At each measurement occasion, the outcome vari-able is explained simultaneously by growth curve factors and byrepeatedly measured self-esteem (i.e., the TVC). Consequently, themodels provide two types of information on the effects of self-esteem: The models estimate (a) the concurrent effect of self-esteem on the outcome while controlling for systematic growth in

the outcome and (b) the expected growth in the outcome whenself-esteem is held constant (Bollen & Curran, 2006; Preacher etal., 2008).

Table 4 reports the results for the TVC models. To test whethercontrolling for self-esteem altered the trajectory of the outcome,we compared the fit of two models. In one model, the growth

Figure 3. Growth curve model of an outcome variable (Y) with self-esteem as a time-varying covariate, shownfor quadratic growth (the models for linear and cubic growth were specified accordingly). The model capturesthe development of Y across the entire observed age range by using individual slope loadings. Parameters withindividually varying values are represented by diamonds (Mehta & West, 2000; Preacher et al., 2008). Linearslope loadings at assessments from 1988 to 2000 are denoted as s1 through s5, and quadratic slope loadings aredenoted as q1 through q5. Individual values for these loadings (i.e., age at assessments and the squared values,respectively; age was centered at 50 years) are included in the analysis through individual data vectors. Residualvariances are denoted as e1 through e5. The model includes covariances between growth factors and theself-esteem scores at assessments from 1988 to 2000.

Table 4Effect of Controlling for Self-Esteem on Life-Span Trajectories of Outcome Variables

Outcome variable

BIC

Self-esteem TVCeffect

Model with growth parametersfreely estimated

Model with growth parametersconstrained to basic model

Relationship satisfaction 24,212.2 24,171.5 0.23�

Job satisfaction 22,750.7 22,737.8 0.26�

Occupational status 21,940.3 21,881.0 0.03Positive affect 29,868.0 29,887.6 0.38�

Negative affect 29,166.4 29,145.6 �0.37�

Depression 28,707.1 28,765.5 �0.52�

Health 29,344.2 29,292.1 0.11�

Note. For BIC, lower values indicate better model fit. The TVC effects of self-esteem are unstandardizedregression coefficients (standardized coefficients are not available). Values in bold indicate the best fittingmodel. BIC � Bayesian information criterion; TVC � time-varying covariate.� p � .05.


parameters were freely estimated, allowing the trajectory to devi-ate from the basic model for this variable. In the other model, thegrowth parameters were fixed to the values from the basic model,assuming that the trajectory is unaltered by controlling for self-esteem. For each outcome variable, we selected the model thatbetter fit the data. The results indicated that controlling for self-esteem altered the trajectory of positive affect and depression butdid not alter trajectories of the other variables. For the modelsselected, we then examined the concurrent TVC effects of self-esteem (Table 4, right column). All effects were in the expecteddirection. The strongest effect of self-esteem emerged for depres-sion: A one-unit increase in self-esteem corresponded to a decreaseof 0.52 units in the expected level of depression. Medium-sizedeffects emerged for positive and negative affect, small to medium-sized effects emerged for relationship satisfaction and job satis-faction, and a very small effect emerged for health. The effect ofself-esteem on the trajectory of occupational status was nonsignif-icant and close to zero.

To examine the results in more detail, we plotted predictedtrajectories (Figure 4). The figure shows the average trajectory foreach outcome variable. If self-esteem had a significant concurrentTVC effect on the outcome variable—which was the case for alloutcome variables except occupational status—the figure also in-cludes trajectories for individuals with constantly high and con-stantly low self-esteem (corresponding to one standard-deviationunit above and below the mean, respectively). If controlling forself-esteem altered the expected growth curve of the outcomevariable—which was the case for positive affect and depression—the figure also includes the controlled trajectory (which is theexpected trajectory for an individual with a constantly averagelevel of self-esteem).

Several findings shown in the figures merit attention. First,the average trajectories of the outcome variables are consistentwith the findings reported in the literature, with the exception ofpositive affect. In this study, positive affect decreased linearlyfrom adolescence to old age (with the difference between age16 and age 97 corresponding to a medium-sized effect, d ��0.58). As reviewed in the introduction, previous researchsuggests that positive affect is relatively stable across adulthoodand decreases slightly only in old age. As stated, the trajectoriesof the other outcome variables converge with findings fromprevious studies. Relationship satisfaction showed a U-shapedtrajectory, decreasing slightly until age 46 (d � �0.23) andthen increasing more strongly into old age (d � 0.68). Jobsatisfaction increased linearly across the observed age range(i.e., age 16 to age 70; d � 0.85). Occupational status increasedstrongly from age 16 to age 46 (d � 1.95) and then decreasedslightly (d � �0.19). Negative affect decreased strongly fromadolescence to middle adulthood and then leveled off in old age(with the difference between age 16 and age 97 correspondingto a very strong effect, d � �1.55). Depression decreased fromadolescence to about age 60 and then increased again into oldage (with the difference between age 16 and age 60 correspond-ing to a medium-to-strong effect, d � �0.60, and the increasebetween age 60 and age 97 corresponding to a strong effect, d �0.83). Finally, health increased slightly from age 16 to age 46(d � 0.09) and then decreased strongly at an accelerating rateinto old age (d � �0.87).

Second, the figures illustrate the size of the self-esteem effectson the level of the outcome trajectories (i.e., the concurrent TVCeffects). As mentioned, the largest effects emerged for positiveaffect, negative affect, and depression, as illustrated by the dis-tance between trajectories for individuals with low versus highself-esteem. In contrast, for relationship satisfaction, job satisfac-tion, and health, the self-esteem effects on the level of the trajec-tories were smaller.

Third, for positive affect and depression, the figures illustratethe effect of controlling for self-esteem on the expected trajectories(Figures 4D and 4F). Whereas the effect is very small for positiveaffect (e.g., at age 97, the difference between the uncontrolled andcontrolled trajectory corresponded to d � �0.03), the effect islarger for depression (at age 97 the difference corresponded to d �0.29). For depression, controlling for self-esteem attenuated thedecline from adolescence to middle adulthood and also attenuatedthe increase from middle adulthood to old age.

Discussion

In the present research, we used data from a large longitudinalstudy to investigate the life-span development of self-esteem andthe prospective influence of self-esteem on life outcomes, includ-ing relationship satisfaction, job satisfaction, occupational status,salary, positive and negative affect, depression, and health. Severalimportant findings emerged. First, we replicated the previouslyreported curvilinear trajectory of self-esteem (Orth et al., 2010);specifically, self-esteem increased from adolescence to middleadulthood, reached a peak at about age 50, and then decreased inold age. Second, self-esteem had significant cross-lagged effectson all of the life outcomes examined in the present research, but noreciprocal effects of life outcomes on self-esteem were found; thispattern is consistent with the hypothesis that self-esteem is a cause,rather than a consequence, of life outcomes (Swann et al., 2007;Trzesniewski et al., 2006). Third, growth curve analyses, withself-esteem as a TVC, suggested that self-esteem has medium-sized effects on the life-span trajectories of affect and depression,small to medium-sized effects on the trajectories of relationshipand job satisfaction, a very small but significant effect on thetrajectory of health, and no significant effect on the trajectory ofoccupational status. Moreover, for some of the outcome variables(specifically, positive affect and depression), self-esteem not onlymoderated the level but also the slope of the trajectories. Next, wediscuss each of these findings in more detail.

Implications of the Findings

The present research replicates in an independent sample thecurvilinear trajectory of self-esteem previously found in analysesof the Americans’ Changing Lives study (Orth et al., 2010; Shawet al., 2010). Although the present study suggests an earlier peakof the self-esteem trajectory (i.e., at about age 50 years) than inprevious research (at about age 60 years; Orth et al., 2010), theoverall shape of the trajectory was similar. The repeated finding ofa relatively strong decline of self-esteem in old age is of particularimportance, given conflicting reviews of the literature (Bengtson,Reedy, & Gordon, 1985; Demo, 1992), conflicting results fromstudies focusing on old age (e.g., Coleman, Ivani-Chalian, &Robinson, 1993; Gove, Ortega, & Style, 1989; Ranzjin, Keeves,


Figure 4. Predicted trajectories of outcome variables. The figure shows average trajectories (continuous lines),trajectories for individuals with constantly high self-esteem (long dash-dot lines), trajectories for individuals withconstantly low self-esteem (dashed lines), and—for positive affect (Panel D) and depression (Panel F) for whichthe trajectory was altered when self-esteem was included as a time-varying covariate—trajectories for individ-uals with a constantly average level of self-esteem (dotted lines). High and low self-esteem corresponded to onestandard-deviation unit above and below the mean, respectively. For job satisfaction and occupational status, theage range is restricted to 16 to 70 years because only a few observations were available outside of this range.For occupational status, only the average trajectory is shown because self-esteem did not significantly predict thetrajectory of this outcome variable. The measures were converted to z scores for the analyses.


Luszcz, & Feather, 1998; Reitzes, Mutran, & Fernandez, 1996),and the negative impact of low self-esteem on a person’s generalwell-being.

We also investigated possible moderators of the life-span tra-jectory of self-esteem. Consistent with previous research (Orth etal., 2010), educational attainment affected the overall level but notthe shape of the trajectory; specifically, the self-esteem trajectoryfor more educated individuals was consistently higher than thetrajectory for less educated individuals, but individuals at bothhigh and low education levels tended to show a curvilinear trend.Surprisingly, gender did not affect the level or the trajectory ofself-esteem; in contrast, previous research has suggested that mentend to report higher self-esteem than women, at least in adoles-cence and adulthood, although the effect size is generally small(Kling, Hyde, Showers, & Buswell, 1999; Orth et al., 2010;Robins, Trzesniewski, Tracy, Gosling, & Potter, 2002). Moreover,in the present study, no cohort differences in the trajectory ofself-esteem were found, replicating findings from Erol and Orth(2011) and Orth et al. (2010). Thus, although the claim that therehas been a generational increase in self-esteem levels (i.e., thatmore recent generations have higher self-esteem than previousgenerations) has intuitive appeal (Twenge & Campbell, 2001,2008), the available evidence suggests that the average self-esteemtrajectory has not changed across the generations born in the 20thcentury (Trzesniewski & Donnellan, 2010; Trzesniewski, Donnel-lan, & Robins, 2008). Finally, it is worth noting that the self-esteem measure used (i.e., the RSE factor scores) showed stronglongitudinal measurement invariance. This result strengthens con-fidence in the validity of the growth curve analyses given thatmeasurement invariance is a fundamental but rarely tested assump-tion on which growth models are based (Edwards & Wirth, 2009;Widaman et al., 2010).

The present research also addressed the important question ofwhether self-esteem is better thought of as a cause or a conse-quence of life outcomes. We tested for reciprocal prospectiverelations between self-esteem and a set of life outcomes that arecentral to having a successful and fulfilling life, including mea-sures of well-being (positive affect, negative affect, and depres-sion), enjoying and succeeding in work, having a satisfying ro-mantic relationship, and physical health. With regard todepression, we replicated previous studies showing that low self-esteem prospectively predicts depression but that the effect ofdepression on low self-esteem is small or nonsignificant (Metalskyet al., 1993; Orth, Robins, & Meier, 2009; Orth, Robins, Trzesni-ewski, et al., 2009; Roberts & Monroe, 1992). A similar patternemerged for measures of dispositional positive and negative affect:Self-esteem predicted increases in positive affect and decreases innegative affect, controlling for prior levels in the constructs, butpositive affect did not predict subsequent self-esteem, and negativeaffect had a statistically significant but small negative effect onself-esteem. In addition, we found that self-esteem was prospec-tively related to higher levels of relationship satisfaction, jobsatisfaction, occupational status, salary, and physical health, con-trolling for prior levels of these variables, but none of these lifeoutcomes had reciprocal effects on self-esteem (or, if significant,the coefficients were small). Moreover, all results held acrossgenerations. Thus, regardless of whether one was born in the early1900s or in the 1980s, self-esteem had significant benefits forpeople’s experiences of love, work, and health, supporting hypoth-

eses about the beneficial consequences of high self-esteem (Don-nellan, Trzesniewski, Robins, Moffitt, & Caspi, 2005; Swann etal., 2007; Trzesniewski et al., 2006; but see Baumeister et al.,2003). In this context, it is worth noting that the subjective expe-rience, or phenomenological state, of having high self-esteem andfeeling positive about oneself is an intrinsically desirable end state,regardless of whether it causes the individual to get better grades,earn more money, live longer, engage in less crime, or achieveother objective outcomes.

Finally, we investigated the influence of self-esteem on thelife-span trajectories of life outcomes, given that the cross-laggedanalyses suggested that self-esteem has prospective effects on theoutcome variables but not vice versa. Whereas the cross-laggedanalyses tested whether self-esteem predicts change in the rank-order position in life outcomes, the growth curve analyses testedwhether self-esteem moderates the life-span trajectory of life out-comes. The results showed that self-esteem had small to medium-sized effects on the life-span trajectories of relationship satisfac-tion, job satisfaction, positive affect, negative affect, anddepression and had a very small effect on the life-span trajectoryof health. In contrast, self-esteem did not significantly predict thetrajectory of a person’s occupational status. As in the growth curveanalyses of self-esteem, we tested for cohort differences in thelife-span trajectories of the outcome variables. For all variables,results suggested that that any existing cohort differences were toosmall to preclude constructing a single overall trajectory fromadolescence to old age. Although some may be surprised that wefound no significant cohort differences for any of the outcomevariables, research in other domains, such as the Big Five person-ality traits, likewise suggest that cohort differences are typicallysmall or nonexistent (Terracciano, McCrae, Brant, & Costa, 2005).

Our results allow us to compare the size of the self-esteemeffects to the size of normative changes in the life outcomes acrossthe life span. For example, the effect of self-esteem on the trajec-tory of relationship satisfaction was 0.23 (indicating that a onestandard-deviation unit increase in self-esteem predicted a 0.23standard-deviation unit increase in relationship satisfaction). Incomparison, relationship satisfaction decreased from age 16 to age46 by 0.23 standard-deviation units and then increased into old ageby 0.68 standard-deviation units. These data indicate that theself-esteem effect and the normative change from adolescence tomidlife are of comparable size; for example, if an individualimproves his or her rank in self-esteem by one standard-deviationunit (relative to other individuals in his age group), the resultingchange in relationship satisfaction outweighs the expected norma-tive loss from age 16 to age 46. In contrast, the average increase inrelationship satisfaction from midlife to old age largely overridesthe moderating effects of self-esteem. Another example is jobsatisfaction: The normative change from age 16 to age 70 corre-sponded to d � 0.85, whereas the self-esteem effect was 0.26;thus, although self-esteem can significantly moderate the trajectoryof job satisfaction, it is unlikely that self-esteem would substan-tially alter the general trend of the job satisfaction trajectory.

Limitations and Future Directions

One limitation is that the sample, although large and economi-cally diverse, is not representative of the population of the UnitedStates. Therefore, future research should test whether the results


hold in other, ideally nationally representative, samples. Moreover,future research should examine the effects of self-esteem on thedevelopment of life outcomes in countries from more diversecultural contexts, such as Asian and African cultures (cf. Arnett,2008). For example, individuals from Asian and Western culturesshow different self-construal styles and different tendencies to-ward self-enhancement (Heine, Lehman, Markus, & Kitayama,1999; Markus & Kitayama, 1991), which may have importantconsequences for the size and even the direction of self-esteemeffects on trajectories of relationship, work, and health outcomes.Therefore, whether studies with samples from other cultural con-texts would yield the same results as the present study is currentlyunknown.

Although we found—in the second part of the study—prospective effects of self-esteem on life outcomes while control-ling for prior levels of the constructs, the study design does notallow for strong conclusions regarding the causal influence ofself-esteem. As in all passive observational designs, effects be-tween variables may be caused by third variables that were notassessed (Finkel, 1995). For example, personality factors, such asneuroticism, might simultaneously affect both self-esteem andmany of the outcome variables. Therefore, future research shouldtest theoretically relevant third-variable models that might accountfor the relations between self-esteem and life outcomes. Neverthe-less, the prospective models (i.e., cross-lagged regression models)are useful because they can indicate whether the data are consistentwith a causal model of the relation between the variables, byestablishing the direction of the effects and ruling out some (butnot all) alternative causal hypotheses. Further evidence on thecausal status of the effects might also accrue from interventionstudies. For example, if improvement of self-esteem through psy-chological intervention were followed by improvements in rela-tionship satisfaction, success at the workplace, psychological well-being, and health, the causal status of self-esteem would beenhanced. Although we do not yet know whether interventions toimprove self-esteem lead to improvements in the relationship,work, and health domains, meta-analytic reviews of self-esteemintervention programs have demonstrated that “it is possible tosignificantly improve children’s and adolescents’ levels of [self-esteem/self-concept] and to obtain concomitant positive changes inother areas of adjustment” (Haney & Durlak, 1998, p. 429; see alsoMarsh & Craven, 2006; O’Mara, Marsh, Craven, & Debus, 2006).

Future research should include informant-based measures andadditional objective measures of life outcomes. This could bedone, for example, in the relationship domain (e.g., partner ratingsof relationship quality), the work domain (e.g., supervisor and peerratings of job performance), and the health domain (e.g., measuresof cardiovascular health, immune functioning, etc.). Typically,correlations between measures that are based on the same method(e.g., self-report) are artificially inflated by shared method vari-ance. In the present context, however, shared method variance isunlikely to account for the cross-lagged effects because some of ithas been statistically removed by controlling for concurrent rela-tions and prior levels of each construct. Moreover, we includedthree objective measures of life outcomes (occupational status,salary, health problems), which showed the same pattern of resultsas the self-report measures. Self-esteem prospectively predictedthe objective outcomes, whereas the prospective effects of theoutcomes on self-esteem were generally nonsignificant. Although

the self-esteem effects were small, they may accumulate to mean-ingful differences over the course of a person’s life. Or, to frameit another way, although the effects might be relatively small foreach individual person, the implications at the societal level couldbe quite large. For example, the effect of aspirin on heart diseaseis .02 (Meyer et al., 2001), yet doctors regularly prescribe aspirinwhen people have heart problems because the reduction in risk ismeaningful at the societal level when millions of people are takingaspirin.

Future research should seek to identify the cognitive, emotional,behavioral, and social processes that mediate the effects of self-esteem on life outcomes. These processes will likely differ acrosslife domains, although some general processes might account foreffects in several life domains. For example, a possible behavioralpathway is that low self-esteem motivates social avoidance andwithdrawal (e.g., Murray et al., 2000), thereby impeding socialreinforcement and social support (Ottenbreit & Dobson, 2004),which likely has negative impact on life domains, such as rela-tionships, work, and health. A possible intrapersonal pathway isthat low self-esteem may increase the tendency to ruminate aboutnegative aspects of the self (Cambron, Acitelli, & Pettit, 2009;Luyckx et al., 2008). Rumination, in turn, strengthens negativeaffect and depression (Mor & Winquist, 2002; Nolen-Hoeksema,2000), which may have further detrimental consequences for jobperformance, relationship quality, and health.

Finally, in future research, it would be interesting to examine theeffects of self-esteem after controlling for narcissism. Althoughself-esteem is only moderately related to narcissism and self-enhancement, in particular when measured by the RSE (Ackermanet al., 2011; Brown & Zeigler-Hill, 2004; Kwan, John, Kenny,Bond, & Robins, 2004; Robins et al., 2001), it is possible that theeffects of self-esteem on life outcomes are even stronger oncenarcissistic self-enhancement is controlled for. In addition, itwould be interesting to test whether the effect of self-esteem onlife outcomes is similar for implicit measures of self-esteem.However, although implicit measures are a promising avenue forself-esteem measurement, there is not yet sufficiently strong sup-port for their validity (Buhrmester, Blanton, & Swann, 2011).

In summary, the present research contributes to the understand-ing of the life-span development of self-esteem and its possibleconsequences for important life outcomes. Our findings are con-sistent with the hypothesis that self-esteem has a significant pro-spective impact on real-world life experiences and that high andlow self-esteem are not mere epiphenomena of success and failurein relevant life domains. An important task in future research is tofurther test whether self-esteem causally influences well-being andsuccess in the domains of work, relationships, and health, forexample, by examining the long-term effects of interventionsaimed at increasing self-esteem.

References

Ackerman, R. A., Witt, E. A., Donnellan, M. B., Trzesniewski, K. H.,Robins, R. W., & Kashy, D. A. (2011). What does the NarcissisticPersonality Inventory really measure? Assessment, 18, 67–87.

Aldwin, C. M., Spiro, A., Levenson, M. R., & Cupertino, A. P. (2001).Longitudinal findings from the Normative Aging Study: III. Personality,individual health trajectories, and mortality. Psychology and Aging, 16,450–465.


Allison, P. D. (2003). Missing data techniques for structural equationmodeling. Journal of Abnormal Psychology, 112, 545–557.

Anderson, S. A., Russell, C. S., & Schumm, W. R. (1983). Perceivedmarital quality and family life-cycle categories: A further analysis.Journal of Marriage and the Family, 45, 127–139.

Andrews, B., & Brown, G. W. (1995). Stability and change in lowself-esteem: The role of psychosocial factors. Psychological Medicine,25, 23–31.

Arnett, J. J. (2008). The neglected 95%: Why American psychology needsto become less American. American Psychologist, 63, 602–614.

Aspinwall, L. G., & Taylor, S. E. (1992). Modeling cognitive adaptation:A longitudinal investigation of the impact of individual differences andcoping on college adjustment and performance. Journal of Personalityand Social Psychology, 63, 989–1003.

Bachman, J. G., & O’Malley, P. M. (1977). Self-esteem in young men: Alongitudinal analysis of the impact of educational and occupationalattainment. Journal of Personality and Social Psychology, 35, 365–380.

Baltes, P. B., Cornelius, S. W., & Nesselroade, J. R. (1979). Cohort effectsin developmental psychology. In J. R. Nesselroade & P. B. Baltes (Eds.),Longitudinal research in the study of behavior and development (pp.61–87). New York, NY: Academic Press.

Baumeister, R. F., Campbell, J. D., Krueger, J. I., & Vohs, K. D. (2003).Does high self-esteem cause better performance, interpersonal success,happiness, or healthier lifestyles? Psychological Science in the PublicInterest, 4, 1–44.

Bengtson, V. L. (2009). Longitudinal Study of Generations, 1971, 1985,1988, 1991, 1994, 1997, 2000 [Data file and codebook]. Ann Arbor, MI:Inter-University Consortium for Political and Social Research [Distrib-utor].

Bengtson, V. L., Biblarz, T. J., & Roberts, R. E. L. (2002). How familiesstill matter: A longitudinal study of youth in two generations. Cam-bridge, United Kingdom: Cambridge University Press.

Bengtson, V. L., Reedy, M. N., & Gordon, C. (1985). Aging and self-conceptions: Personality processes and social contexts. In J. E. Birren &K. W. Schaie (Eds.), Handbook of the psychology of aging (pp. 544–593). New York, NY: Van Nostrand Reinhold.

Benyamini, Y., Leventhal, H., & Leventhal, E. A. (2004). Self-rated oralhealth as an independent predictor of self-rated general health, self-esteem and life satisfaction. Social Science and Medicine, 59, 1109–1116.

Bernal, D., Snyder, D., & McDaniel, M. (1998). The age and job satisfac-tion relationship: Does its shape and strength still evade us? Journal ofGerontology: Psychological Sciences, 53B, P287–P293.

Blanchflower, D. G., & Oswald, A. J. (2008). Is well-being U-shaped overthe life cycle? Social Science and Medicine, 66, 1733–1749.

Boden, J. M., Fergusson, D. M., & Horwood, L. J. (2008). Does adolescentself-esteem predict later life outcomes? A test of the causal role ofself-esteem. Development and Psychopathology, 20, 319–339.

Bollen, K. A., & Curran, P. J. (2006). Latent curve models: A structuralequation perspective. Hoboken, NJ: Wiley.

Bollen, K. A., & Lennox, R. (1991). Conventional wisdom on measure-ment: A structural equation perspective. Psychological Bulletin, 110,305–314.

Boyd, M. (2008). A socioeconomic scale for Canada: Measuring occupa-tional status from the census. Canadian Review of Sociology, 45, 51–91.

Bradburn, N. M. (1969). The structure of psychological well-being. Chi-cago, IL: Aldine.

Brown, R. P., & Zeigler-Hill, V. (2004). Narcissism and the non-equivalence of self-esteem measures: A matter of dominance? Journal ofResearch in Personality, 38, 585–592.

Buhrmester, M. D., Blanton, H., & Swann, W. B. (2011). Implicit self-esteem: Nature, measurement, and a new way forward. Journal ofPersonality and Social Psychology, 100, 365–385.

Cambron, M. J., Acitelli, L. K., & Pettit, J. W. (2009). Explaining gender

differences in depression: An interpersonal contingent self-esteem per-spective. Sex Roles, 61, 751–761.

Carstensen, L. L., Pasupathi, M., Mayr, U., & Nesselroade, J. R. (2000).Emotional experience in everyday life across the adult life span. Journalof Personality and Social Psychology, 79, 644–655.

Charles, S. T. (2010). Strength and vulnerability integration: A model ofemotional well-being across adulthood. Psychological Bulletin, 136,1068–1091.

Charles, S. T., Reynolds, C. A., & Gatz, M. (2001). Age-related differencesand change in positive and negative affect over 23 years. Journal ofPersonality and Social Psychology, 80, 136–151.

Chiu, C. J., & Wray, L. A. (2010). Physical disability trajectories in olderAmericans with and without diabetes: The role of age, gender, race orethnicity, and education. Gerontologist, 51, 51–63.

Christie-Mizell, C. A., Ida, A. K., & Keith, V. M. (2010). African Amer-icans and physical health: The consequences of self-esteem and happi-ness. Journal of Black Studies, 40, 1189–1211.

Clark, A., Oswald, A., & Warr, P. (1996). Is job satisfaction U-shapedin age? Journal of Occupational and Organizational Psychology, 69,57– 81.

Cole, D. A., & Maxwell, S. E. (2003). Testing mediational models withlongitudinal data: Questions and tips in the use of structural equationmodeling. Journal of Abnormal Psychology, 112, 558–577.

Coleman, P. G., Ivani-Chalian, C., & Robinson, M. (1993). Self-esteemand its sources: Stability and change in later life. Ageing and Society, 13,171–192.

Davey, A., Halverson, C. F., Zonderman, A. B., & Costa, P. T. (2004).Change in depressive symptoms in the Baltimore Longitudinal Study ofAging. Journal of Gerontology: Psychological Sciences, 59B, P270–P277.

Demo, D. H. (1992). The self-concept over time: Research issues anddirections. Annual Review of Sociology, 18, 303–326.

Donnellan, M. B., Trzesniewski, K. H., Robins, R. W., Moffitt, T. E., &Caspi, A. (2005). Low self-esteem is related to aggression, antisocialbehavior, and delinquency. Psychological Science, 16, 328–335.

Duncan, T. E., Duncan, S. C., & Strycker, L. A. (2006). An introduction tolatent variable growth curve modeling: Concepts, issues, and applica-tions. Mahwah, NJ: Erlbaum.

Eaton, W. W., Smith, C., Ybarra, M., Muntaner, C., & Tien, A. (2004).Center for Epidemiologic Studies Depression Scale: Review and revi-sion (CESD and CESD-R). In M. E. Maruish (Ed.), The use of psycho-logical testing for treatment planning and outcomes assessment: Vol. 3.Instruments for adults (pp. 363–377). Mahwah, NJ: Erlbaum.

Edwards, M. C., & Wirth, R. J. (2009). Measurement and the study ofchange. Research in Human Development, 6, 74–96.

Elstad, J. I. (2004). Health and status attainment: Effects of health onoccupational achievement among employed Norwegian men. Acta So-ciologica, 47, 127–140.

Erol, R. Y., & Orth, U. (2011). Self-esteem development from age 14 to 30years: A longitudinal study. Journal of Personality and Social Psychol-ogy, 101, 607–619.

Finkel, S. E. (1995). Causal analysis with panel data. Thousand Oaks, CA:Sage.

Ganzeboom, H. B. G., De Graaf, P. M., Treiman, D. J., & de Leeuw, J.(1992). A standard international socio-economic index of occupationalstatus. Social Science Research, 21, 1–56.

Gatz, M., & Hurwicz, M. L. (1990). Are old people more depressed?Cross-sectional data on Center for Epidemiological Studies DepressionScale factors. Psychology and Aging, 5, 284–290.

Gilford, R., & Bengtson, V. (1979). Measuring marital satisfaction in threegenerations: Positive and negative dimensions. Journal of Marriage andthe Family, 41, 387–398.

Gorchoff, S. M., John, O. P., & Helson, R. (2008). Contextualizing change


in marital satisfaction during middle age: An 18-year longitudinal study.Psychological Science, 19, 1194–1200.

Gove, W. R., Ortega, S. T., & Style, C. B. (1989). The maturational androle perspectives on aging and self through the adult years: An empiricalevaluation. American Journal of Sociology, 94, 1117–1145.

Grimm, K. J. (2007). Multivariate longitudinal methods for studying de-velopmental relationships between depression and academic achieve-ment. International Journal of Behavioral Development, 31, 328–339.

Gross, J. J., Carstensen, L. L., Pasupathi, M., Tsai, J., Skorpen, C. G., &Hsu, A. Y. C. (1997). Emotion and aging: Experience, expression, andcontrol. Psychology and Aging, 12, 590–599.

Haney, P., & Durlak, J. A. (1998). Changing self-esteem in children andadolescents: A meta-analytic review. Journal of Clinical Child Psychol-ogy, 27, 423–433.

Harding, S. D. (1982). Psychological well-being in Great Britain: Anevaluation of the Bradburn Affect Balance Scale. Personality and Indi-vidual Differences, 3, 167–175.

Hauser, R. M., Sheridan, J. T., & Warren, J. R. (1999). Socioeconomicachievements of siblings in the life course: New findings from theWisconsin Longitudinal Study. Research on Aging, 21, 338–378.

Hauser, R. M., & Warren, J. R. (1997). Socioeconomic indexes for occu-pations: A review, update, and critique. Sociological Methodology, 27,177–298.

Heine, S. J., Lehman, D. R., Markus, H. R., & Kitayama, S. (1999). Is therea universal need for positive self-regard? Psychological Review, 106,766–794.

Helson, R., & Soto, C. J. (2005). Up and down in middle age: Monotonicand nonmonotonic changes in roles, status, and personality. Journal ofPersonality and Social Psychology, 89, 194–204.

Hochwarter, W. A., Ferris, G. R., Perrewe, P. L., Witt, L. A., & Kiewitz,C. (2006). A note on the nonlinearity of the age–job-satisfaction rela-tionship. Journal of Applied Social Psychology, 31, 1223–1237.

House, J. S., Lepkowski, J. M., Kinney, A. M., Mero, R. P., Kessler, R. C.,& Herzog, A. R. (1994). The social stratification of aging and health.Journal of Health and Social Behavior, 35, 213–234.

Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariancestructure analysis: Conventional criteria versus new alternatives. Struc-tural Equation Modeling, 6, 1–55.

Huang, C. (2010). Mean-level change in self-esteem from childhoodthrough adulthood: Meta-analysis of longitudinal studies. Review ofGeneral Psychology, 14, 251–260.

Hunt, J. W., & Saul, P. N. (1975). The relationship of age, tenure, and jobsatisfaction in males and females. Academy of Management Journal, 18,690–702.

Janson, P., & Martin, J. K. (1982). Job satisfaction and age: A test of twoviews. Social Forces, 60, 1089–1102.

Joiner, T. E. (1995). The price of soliciting and receiving negative feed-back: Self-verification theory as a vulnerability to depression theory.Journal of Abnormal Psychology, 104, 364–372.

Judge, T. A., & Bono, J. E. (2001). Relationship of core self-evaluationstraits—self-esteem, generalized self-efficacy, locus of control, and emo-tional stability—with job satisfaction and job performance: A meta-analysis. Journal of Applied Psychology, 86, 80–92.

Judge, T. A., Bono, J. E., & Locke, E. A. (2000). Personality and jobsatisfaction: The mediating role of job characteristics. Journal of AppliedPsychology, 85, 237–249.

Judge, T. A., & Hurst, C. (2008). How the rich (and happy) get richer (andhappier): Relationship of core self-evaluations to trajectories in attainingwork success. Journal of Applied Psychology, 93, 849–863.

Judge, T. A., Hurst, C., & Simon, L. S. (2009). Does it pay to be smart,attractive, or confident (or all three)? Relationships among generalmental ability, physical attractiveness, core self-evaluations, and in-come. Journal of Applied Psychology, 94, 742–755.

Kalleberg, A. L., & Loscocco, K. A. (1983). Aging, values, and rewards:

Explaining age differences in job satisfaction. American SociologicalReview, 48, 78–90.

Kammeyer-Mueller, J. D., Judge, T. A., & Piccolo, R. F. (2008). Self-esteem and extrinsic career success: Test of a dynamic model. AppliedPsychology: An International Review, 57, 204–224.

Karney, B. R., & Bradbury, T. N. (1995). The longitudinal course ofmarital quality and stability: A review of theory, method, and research.Psychological Bulletin, 118, 3–34.

Kasen, S., Cohen, P., Chen, H., & Castille, D. (2003). Depression in adultwomen: Age changes and cohort effects. American Journal of PublicHealth, 93, 2061–2066.

Kessler, E. M., & Staudinger, U. M. (2009). Affective experience inadulthood and old age: The role of affective arousal and perceived affectregulation. Psychology and Aging, 24, 349–362.

Kessler, R. C., Foster, C., Webster, P. S., & House, J. S. (1992). Therelationship between age and depressive symptoms in two nationalsurveys. Psychology and Aging, 7, 119–126.

Kim, J., & Miech, R. (2009). The Black–White difference in age trajec-tories of functional health over the life course. Social Science andMedicine, 68, 717–725.

Kim, K. A., & Mueller, D. J. (2001). To balance or not to balance:Confirmatory factor analysis of the Affect-Balance Scale. Journal ofHappiness Studies, 2, 289–306.

Kling, K. C., Hyde, J. S., Showers, C. J., & Buswell, B. N. (1999). Genderdifferences in self-esteem: A meta-analysis. Psychological Bulletin, 125,470–500.

Krueger, J. I., Vohs, K. D., & Baumeister, R. F. (2008). Is the allure ofself-esteem a mirage after all? American Psychologist, 63, 64–65.

Kunzmann, U., Little, T. D., & Smith, J. (2000). Is age-related stability ofsubjective well-being a paradox? Cross-sectional and longitudinal evi-dence from the Berlin Aging Study. Psychology and Aging, 15, 511–526.

Kwan, V. S. Y., John, O. P., Kenny, D. A., Bond, M. H., & Robins, R. W.(2004). Reconceptualizing individual differences in self-enhancementbias: An interpersonal approach. Psychological Review, 111, 94–110.

Leary, M. R., & Baumeister, R. F. (2000). The nature and function ofself-esteem: Sociometer theory. In M. P. Zanna (Ed.), Advances inexperimental social psychology (Vol. 32, pp. 1–62). San Diego, CA:Academic Press.

Lewinsohn, P. M., Rohde, P., Seeley, J. R., & Fischer, S. A. (1991). Ageand depression: Unique and shared effects. Psychology and Aging, 6,247–260.

Liang, J., Shaw, B. A., Krause, N. M., Bennett, J. M., Blaum, C., Ko-bayashi, E., . . . Sugisawa, H. (2003). Changes in functional statusamong older adults in Japan: Successful and usual aging. Psychologyand Aging, 18, 684–695.

Liang, J., Shaw, B. A., Krause, N. M., Bennett, J. M., Kobayashi, E.,Fukaya, T., & Sugihara, Y. (2005). How does self-assessed healthchange with age? A study of older adults in Japan. Journal of Geron-tology: Social Sciences, 60B, S224–S232.

Little, T. D., Preacher, K. J., Selig, J. P., & Card, N. A. (2007). Newdevelopments in latent variable panel analyses of longitudinal data.International Journal of Behavioral Development, 31, 357–365.

Löckenhoff, C. E., Costa, P. T., & Lane, R. D. (2008). Age differences indescriptions of emotional experiences in oneself and others. Journal ofGerontology: Psychological Sciences, 63B, P92–P99.

Luyckx, K., Schwartz, S. J., Berzonsky, M. D., Soenens, B., Vansteenkiste,M., Smits, I., & Goossens, L. (2008). Capturing ruminative exploration:Extending the four-dimensional model of identity formation in lateadolescence. Journal of Research in Personality, 42, 58–82.

MacCallum, R. C., & Austin, J. T. (2000). Applications of structuralequation modeling in psychological research. Annual Review of Psychol-ogy, 51, 201–226.

MacCallum, R. C., Browne, M. W., & Cai, L. (2006). Testing differences


between nested covariance structure models: Power analysis and nullhypotheses. Psychological Methods, 11, 19–35.

Makikangas, A., Kinnunen, U., & Feldt, T. (2004). Self-esteem, disposi-tional optimism, and health: Evidence from cross-lagged data on em-ployees. Journal of Research in Personality, 38, 556–575.

Markus, H. R., & Kitayama, S. (1991). Culture and the self: Implicationsfor cognition, emotion, and motivation. Psychological Review, 98, 224–253.

Marsh, H. W., & Craven, R. G. (2006). Reciprocal effects of self-conceptand performance from a multidimensional perspective: Beyond seduc-tive pleasure and unidimensional perspectives. Perspectives on Psycho-logical Science, 1, 133–163.

Marsh, H. W., Scalas, L. F., & Nagengast, B. (2010). Longitudinal tests ofcompeting factor structures for the Rosenberg Self-Esteem Scale: Traits,ephemeral artifacts, and stable response styles. Psychological Assess-ment, 22, 366–381.

McCullough, M. E., & Laurenceau, J. P. (2005). Religiousness and thetrajectory of self-rated health across adulthood. Personality and SocialPsychology Bulletin, 31, 560–573.

Mehta, P. D., & West, S. G. (2000). Putting the individual back intoindividual growth curves. Psychological Methods, 5, 23–43.

Metalsky, G. I., Joiner, T. E., Hardin, T. S., & Abramson, L. Y. (1993).Depressive reactions to failure in a naturalistic setting: A test of thehopelessness and self-esteem theories of depression. Journal of Abnor-mal Psychology, 102, 101–109.

Meyer, G. J., Finn, S. E., Eyde, L. D., Kay, G. G., Moreland, K. L., Dies,R. R., . . . Reed, G. M. (2001). Psychological testing and psychologicalassessment: A review of evidence and issues. American Psychologist,56, 128–165.

Miech, R. A., Eaton, W., & Liang, K. Y. (2003). Occupational stratificationover the life course: A comparison of occupational trajectories acrossrace and gender during the 1980s and 1990s. Work and Occupations, 30,440–473.

Miech, R. A., & Shanahan, M. J. (2000). Socioeconomic status anddepression over the life course. Journal of Health and Social Behavior,41, 162–176.

Mirowsky, J., & Kim, J. (2007). Graphing age trajectories: Vector graphs,synthetic and virtual cohort projections, and cross-sectional profiles ofdepression. Sociological Methods and Research, 35, 497–541.

Mirowsky, J., & Reynolds, J. R. (2000). Age, depression, and attrition inthe National Survey of Families and Households. Sociological Methodsand Research, 28, 476–504.

Mirowsky, J., & Ross, C. E. (1992). Age and depression. Journal of Healthand Social Behavior, 33, 187–205.

Mor, N., & Winquist, J. (2002). Self-focused attention and negative affect:A meta-analysis. Psychological Bulletin, 128, 638–662.

Mroczek, D. K. (2001). Age and emotion in adulthood. Current Directionsin Psychological Science, 10, 87–90.

Mroczek, D. K., & Kolarz, C. M. (1998). The effect of age on positive andnegative affect: A developmental perspective on happiness. Journal ofPersonality and Social Psychology, 75, 1333–1349.

Murray, S. L., Holmes, J. G., & Griffin, D. W. (2000). Self-esteem and thequest for felt security: How perceived regard regulates attachment pro-cesses. Journal of Personality and Social Psychology, 78, 478–498.

Murray, S. L., Rose, P., Bellavia, G. M., Holmes, J. G., & Kusche, A. G.(2002). When rejection stings: How self-esteem constrains relationship-enhancing processes. Journal of Personality and Social Psychology, 83,556–573.

Muthen, L. K., & Muthen, B. O. (2010). Mplus user’s guide: Sixth edition.Los Angeles, CA: Muthen & Muthen.

Nam, C. B., & Boyd, M. (2004). Occupational status in 2000: Over acentury of census-based measurement. Population Research and PolicyReview, 23, 327–358.

Ng, T. W. H., & Feldman, D. C. (2010). The relationships of age with jobattitudes: A meta-analysis. Personnel Psychology, 63, 677–718.

Nolen-Hoeksema, S. (2000). The role of rumination in depressive disordersand mixed anxiety/depressive symptoms. Journal of Abnormal Psychol-ogy, 109, 504–511.

O’Mara, A. J., Marsh, H. W., Craven, R. G., & Debus, R. L. (2006). Doself-concept interventions make a difference? A synergistic blend ofconstruct validation and meta-analysis. Educational Psychologist, 41,181–206.

Orbuch, T. L., House, J. S., Mero, R. P., & Webster, P. S. (1996). Maritalquality over the life course. Social Psychology Quarterly, 59, 162–171.

Ormel, J., Oldehinkel, A. J., & Vollebergh, W. (2004). Vulnerabilitybefore, during, and after a major depressive episode. Archives of GeneralPsychiatry, 61, 990–996.

Orth, U., Robins, R. W., & Meier, L. L. (2009). Disentangling the effectsof low self-esteem and stressful events on depression: Findings fromthree longitudinal studies. Journal of Personality and Social Psychology,97, 307–321.

Orth, U., Robins, R. W., & Roberts, B. W. (2008). Low self-esteemprospectively predicts depression in adolescence and young adulthood.Journal of Personality and Social Psychology, 95, 695–708.

Orth, U., Robins, R. W., Trzesniewski, K. H., Maes, J., & Schmitt, M.(2009). Low self-esteem is a risk factor for depressive symptoms fromyoung adulthood to old age. Journal of Abnormal Psychology, 118,472–478.

Orth, U., Trzesniewski, K. H., & Robins, R. W. (2010). Self-esteemdevelopment from young adulthood to old age: A cohort-sequentiallongitudinal study. Journal of Personality and Social Psychology, 98,645–658.

Ottenbreit, N. D., & Dobson, K. S. (2004). Avoidance and depression: Theconstruction of the Cognitive–Behavioral Avoidance Scale. BehaviourResearch and Therapy, 42, 293–313.

Preacher, K. J., Wichman, A. L., MacCallum, R. C., & Briggs, N. E.(2008). Latent growth curve modeling. Los Angeles, CA: Sage.

Radloff, L. S. (1977). The CES-D Scale: A self-report depression scale forresearch in the general population. Applied Psychological Measurement,1, 385–401.

Ralph, J. A., & Mineka, S. (1998). Attributional style and self-esteem: Theprediction of emotional distress following a midterm exam. Journal ofAbnormal Psychology, 107, 203–215.

Ranzjin, R., Keeves, J., Luszcz, M., & Feather, N. T. (1998). The role ofself-perceived usefulness and competence in the self-esteem of elderlyadults: Confirmatory factor analyses of the Bachman revision of Rosen-berg’s Self-Esteem Scale. Journals of Gerontology: Psychological Sci-ences and Social Sciences, 53B, P96–P104.

Raykov, T., Dimitrov, D. M., & Asparouhov, T. (2010). Evaluation ofscale reliability with binary measures using latent variable modeling.Structural Equation Modeling, 17, 265–279.

Reitzes, D. C., & Mutran, E. J. (2006). Self and health: Factors thatencourage self-esteem and functional health. Journal of Gerontology:Social Sciences, 61B, S44–S51.

Reitzes, D. C., Mutran, E. J., & Fernandez, M. E. (1996). Does retirementhurt well-being? Factors influencing self-esteem and depression amongretirees and workers. Gerontologist, 36, 649–656.

Rhodes, S. R. (1983). Age-related differences in work attitudes and be-havior: A review and conceptual analysis. Psychological Bulletin, 93,328–367.

Roberts, J. E., & Monroe, S. M. (1992). Vulnerable self-esteem anddepressive symptoms: Prospective findings comparing three alternativeconceptualizations. Journal of Personality and Social Psychology, 62,804–812.

Robins, R. W., Hendin, H. M., & Trzesniewski, K. H. (2001). Measuringglobal self-esteem: Construct validation of a single-item measure and the


Rosenberg Self-Esteem Scale. Personality and Social Psychology Bul-letin, 27, 151–161.

Robins, R. W., Trzesniewski, K. H., Tracy, J. L., Gosling, S. D., & Potter,J. (2002). Global self-esteem across the life span. Psychology and Aging,17, 423–434.

Rosenberg, M. (1965). Society and the adolescent self-image. Princeton,NJ: Princeton University Press.

Ross, C. E., & Wu, C. L. (1996). Education, age, and the cumulativeadvantage in health. Journal of Health and Social Behavior, 37, 104–120.

Rothermund, K., & Brandtstadter, J. (2003). Depression in later life:Cross-sequential patterns and possible determinants. Psychology andAging, 18, 80–90.

Sacker, A., Worts, D., & McDonough, P. (2011). Social influences ontrajectories of self-rated health: Evidence from Britain, Germany, Den-mark and the USA. Journal of Epidemiology and Community Health, 65,130–136.

Salmela-Aro, K., & Nurmi, J. E. (2007). Self-esteem during universitystudies predicts career characteristics 10 years later. Journal of Voca-tional Behavior, 70, 463–477.

Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the stateof the art. Psychological Methods, 7, 147–177.

Scheibe, S., & Carstensen, L. L. (2010). Emotional aging: Recent findingsand future trends. Journal of Gerontology: Psychological Sciences, 65B,135–144.

Seligman, M. E. P. (1993). What you can change and what you can’t: Thecomplete guide to successful self-improvement. New York, NY: Fawcett.

Shackelford, T. K. (2001). Self-esteem in marriage. Personality and Indi-vidual Differences, 30, 371–390.

Shahar, G., & Davidson, L. (2003). Depressive symptoms erode self-esteem in severe mental illness: A three-wave, cross-lagged study.Journal of Consulting and Clinical Psychology, 71, 890–900.

Shahar, G., & Henrich, C. C. (2010). Do depressive symptoms erodeself-esteem in early adolescence? Self and Identity, 9, 403–415.

Shaw, B. A., Liang, J., & Krause, N. (2010). Age and race differences inthe trajectory of self-esteem. Psychology and Aging, 25, 84–94.

Stinson, D. A., Logel, C., Zanna, M. P., Holmes, J. G., Cameron, J. J.,Wood, J. V., & Spencer, S. J. (2008). The cost of lower self-esteem:Testing a self- and social-bonds model of health. Journal of Personalityand Social Psychology, 94, 412–428.

Streiner, D. L. (2003). Being inconsistent about consistency: When coef-ficient alpha does and doesn’t matter. Journal of Personality Assess-ment, 80, 217–222.

Swann, W. B., Chang-Schneider, C., & McClarty, K. L. (2007). Dopeople’s self-views matter? American Psychologist, 62, 84–94.

Swann, W. B., Chang-Schneider, C., & McClarty, K. L. (2008). Yes,cavalier attitudes can have pernicious consequences. American Journalof Community Psychology, 63, 65–66.

Terracciano, A., McCrae, R. R., Brant, L. J., & Costa, P. T. (2005).Hierarchical linear modeling analyses of the NEO-PI-R scales in theBaltimore Longitudinal Study of Aging. Psychology and Aging, 20,493–506.

Trzesniewski, K. H., & Donnellan, M. B. (2010). Rethinking “Generation

Me”: A study of cohort effects from 1976–2006. Perspectives on Psy-chological Science, 5, 58–75.

Trzesniewski, K. H., Donnellan, M. B., Moffitt, T. E., Robins, R. W.,Poulton, R., & Caspi, A. (2006). Low self-esteem during adolescencepredicts poor health, criminal behavior, and limited economic prospectsduring adulthood. Developmental Psychology, 42, 381–390.

Trzesniewski, K. H., Donnellan, M. B., & Robins, R. W. (2008). Dotoday’s young people really think they are so extraordinary? An exam-ination of secular trends in narcissism and self-enhancement. Psycho-logical Science, 19, 181–188.

Twenge, J. M., & Campbell, W. K. (2001). Age and birth cohort differ-ences in self-esteem: A cross-temporal meta-analysis. Personality andSocial Psychology Review, 5, 321–344.

Twenge, J. M., & Campbell, W. K. (2002). Self-esteem and socioeconomicstatus: A meta-analytic review. Personality and Social Psychology Re-view, 6, 59–71.

Twenge, J. M., & Campbell, W. K. (2008). Increases in positive self-viewsamong high school students: Birth cohort changes in anticipated perfor-mance, self-satisfaction, self-liking, and self-competence. PsychologicalScience, 19, 1082–1086.

Umberson, D., Williams, K., Powers, D. A., Chen, M. D., & Campbell,A. M. (2005). As good as it gets? A life course perspective on maritalquality. Social Forces, 84, 493–511.

Vaillant, C. O., & Vaillant, G. E. (1993). Is the U-curve of maritalsatisfaction an illusion? A 40-year study of marriage. Journal of Mar-riage and the Family, 55, 230–239.

VanLaningham, J., Johnson, D. R., & Amato, P. (2001). Marital happiness,marital duration, and the U-shaped curve: Evidence from a five-wavepanel study. Social Forces, 78, 1313–1341.

Voss, K., Markiewicz, D., & Doyle, A. B. (1999). Friendship, marriage,and self-esteem. Journal of Social and Personal Relationships, 16,103–122.

Wallace, J., & O’Hara, M. W. (1992). Increases in depressive symptom-atology in the rural elderly: Results from a cross-sectional and longitu-dinal study. Journal of Abnormal Psychology, 101, 398–404.

Warr, P. (1992). Age and occupational well-being. Psychology and Aging,7, 37–45.

Watson, D., Suls, J., & Haig, J. (2002). Global self-esteem in relation tostructural models of personality and affectivity. Journal of Personalityand Social Psychology, 83, 185–197.

Widaman, K. F. (2006). Missing data: What to do with or without them.Monographs of the Society for Research in Child Development, 71,42–64.

Widaman, K. F., Ferrer, E., & Conger, R. D. (2010). Factorial invariancewithin longitudinal structural equation models: Measuring the sameconstruct across time. Child Development Perspectives, 4, 10–18.

Wirth, R. J., & Edwards, M. C. (2007). Item factor analysis: Currentapproaches and future directions. Psychological Methods, 12, 58–79.

Received October 17, 2010Revision received July 21, 2011

Accepted August 22, 2011 �


life-span development of self-esteem and its effects on ... · personality processes and individual...

Documents